The Uncertainty of Measurements: Physical and Chemical Metrology: Impact and Analysis

1

Metrology and the Need for Reliable Measurements

Thus every sort of confusion is revealed within us, and the art of measuring, numbering and weighing come into the rescue of human understanding—there is the beauty of them—the apparent greater or less or more or heavier no longer have mystery over us, but give way to measurement and calculation and weight. . . .

Then that part of the soul which has an opinion contrary to measure is not the same with that which has an opinion in accordance with measure. . . .

And better part of the soul is likely to be with that which trusts to measure and calculation. . . .

And that which is opposed to them is one of the inferior principles of the soul.

Socrates as quoted in Plato’s Republic

A measurement is a kind of language. It gives everyone something to agree upon so that effective communication and commerce can take place. Good measurements must be defined by their accuracy, precision and need. Of course, measurement capabilities should fit the need, but new measurement capabilities can often redefine those needs.

David Layden, cited in Quality Progress,

February 2001

An almost infinite number of measurements are made every day in trade, industry, and scientific laboratories throughout the world. The measurement results are used to make decisions which could be crucial at times. These decisions affect our lives as professionals and also as consumers. In order for us to have adequate confidence in the correctness of decisions made based on measurement results, it is essential that the measurement results are reliable. This chapter explains the critical role played by measurements in various activities associated with human beings. The importance of reliable measurements in decision-making and the need for correct understanding and interpretation of measurement results are also explained.

THE UNIVERSAL NATURE OF MEASUREMENTS

The false diagnosis of an ailment may have serious consequences for an individual. Sometimes it can also be very embarrassing for decision makers at the senior level. One such case was reported by the Hindustan Times, a leading English-language daily newspaper published in New Delhi, in its edition of 15 January 2002. A soldier was tested for HIV along with other troops in his regiment when the regiment was selected for duty with a UN peace mission abroad. The soldier was diagnosed as HIV positive, and his medical category was downgraded. This decision was made based on the HIV test carried out in the Armed Forces Transfusion Centre. Feeling humiliated, the soldier had his HIV test conducted in a civil hospital, where he tested negative. The soldier then sued the army authorities for false diagnosis in a court of law. The court issued notice to the army for the facts of the case, and the army authorities filed an affidavit before the court.

In defending their case, the army authorities based their arguments on the laws of probability and on medical literature to explain how the army had diagnosed a soldier as HIV positive when he was in fact HIV negative. They quoted an article by an expert which stated that HIV antibody tests are 99.5% accurate. Cases which are actually HIV negative but test HIV positive on some occasion are termed false positive, and they are not unknown in the medical literature. The army therefore deduced that there is a probability of 1 in 200 persons’ testing false positive. The army authorities also stated that the soldier had again been found HIV positive in the same army center where he had tested HIV positive earlier, even after he tested negative in the civil hospital. For confirmation, the soldier was again tested in another reputed civil hospital and was found HIV negative. The army informed the court that based on the test the soldier was declared HIV negative. The soldier has, however, sought compensation for the humiliation and disgrace that he faced due to the false diagnosis.

Such cases of false diagnosis do occur, though very rarely. It is not uncommon for two laboratories to report different results. In such a situation, it is important to decide whether the difference is because of the inherent variability of the analytical processes at the two laboratories or because of the bias in the analytical process of one of the laboratories. Due to variability, analytical results are probabilistic in nature. There are risks of making a wrong decision based on these results, though the probability of this may be quite low. The preceding example, however, shows what the consequences of such a wrong decision could be.

Effective decision making requires the appropriate use of available measurement data. In this context, the author had an interesting encounter with a colleague about two decades back. The colleague had been working as a technician in a repair and maintenance establishment. One of the job functions assigned to him was measuring certain technical parameters of purchased items and submitting the results to his supervisor for a decision on the acceptance or rejection of the items. He had a son studying in a secondary school. While teaching his son about the Pythagorean theorem in geometry, he had the noble idea of experimentally verifying the statement of the theorem. The next day, while narrating his teaching experience to some of his colleagues, he informed them about his newfound amusing discovery that the statement of Pythagoras’s theorem is not correct. In response to a specific query about the basis of his bold statement, he offered an interesting explanation. In the process of verifying the statement of the theorem, he had drawn a right-angled triangle and measured the lengths of the hypotenuse and the other two sides, and he had observed that the square of the hypotenuse was not exactly equal to the sum of the squares of the other two sides. He repeated the exercise a number of times and drew the same conclusion. It took a lot of effort to make him understand that the exact equivalence could not be achieved due to the presence of a number of factors that contribute to errors in measurement. Some of the important factors in this case could have been the limited resolution of the measuring scale and goniometer, their inherent inaccuracies, and his own level of measuring skill.

Things have not changed too much in the last two decades. One can always find persons who have this type of conceptual misunderstanding about measurement results. They conceive of measurement results as exact and unambiguous. They are ignorant of the fact that measurement results are probabilistic in nature. This implies that another set of measurements of the same parameter could result in values which are not exactly the same as but are close to earlier reported values. It is therefore essential that such misconceptions about measurement results should be guarded against.

Measurements are extensively in demand but often go unnoticed. P. H. Sydenham has reported, based on a survey by Huntoon, that in 1967 about 20 billion measurements were performed each day in the United States. In 1965, the U.S. industrial sector invested around 3% of GNP in measurement. Measurement affects human lives in many ways. We are concerned about day-to-day measurements in the areas of trade, healthcare, and environmental pollution. As individuals, we are faced with measurements taken on a day-to-day basis by market scales; gas, electricity, and water meters; gasoline pumps; taxi meters; and so on. We are also concerned with the reliability of measurements in healthcare, such as the measurement of the temperature of the human body by clinical thermometers, the monitoring of blood pressures and heartbeats, and the clinical analysis of body fluids in pathology laboratories. As a society, we are concerned with the reliability of the measurement results of environmental parameters, such as sulfur dioxide, carbon monoxide, and carbon dioxide concentrations in the atmosphere and the density of suspended particles in the atmosphere.

Following are some examples of how measurement affects our day-today lives.

Trade and consumer protection: A well-defined and accepted measurement system facilitates trade at the local, national, and international levels and protects consumers from unethical trade practices.

Safety and healthcare: Human beings need to be assured of their safety when they use industrial products. This requires the evaluation of risks due to the products’ use as well as the design and manufacture of safe products. At various stages of the manufacturing and service processes, measurements of product parameters that have bearing on human safety are made. Similarly, healthcare involves analytical and clinical measurements for diagnostic purposes in order to decide the course of treatment.

Environmental protection: The environment around us has a direct bearing on the quality of our lives. Because of this, the quality of the air, water, and soil must be monitored regularly. To save the environment, the quality of industrial waste has to be controlled. There are environmental protection laws in most of countries which make it obligatory for individuals and organizations not to pollute the air, water, and soil to dangerous levels. Measurements help in achieving the desired quality level of the environment.

Law and order: Police radar is used to measure the speed of motor vehicles, which helps in maintaining orderly traffic on the roads. The analytical methods used in forensic science laboratories help in solving crimes. The results obtained are presented in legal proceedings as evidence in a court of law. This ensures fair legal judgments.

Measurements are also crucially important for quality professionals working in manufacturing and service organizations. In industrial manufacturing and service organizations, measurements are required to be made at various stages of the life cycle of a product. A number of quality management standards contain specific requirements for the quality of measurement results and their evaluation. These include the ISO 9000 standards for quality management systems, the ISO 14000 standards for environmental management systems, the QS-9000 quality system standards for automotive parts, and the TL 9000 quality standards for the telecommunications industry.

In laboratories, measurements are made of various products to test their characteristics. At the international level, mutual acceptability of the measurements and test results of products is a matter of concern for eliminating technical barriers to trade.

Metrology, the science of measurement, encompasses both the theoretical and the practical aspects of measurement. In spite of the all-pervasive nature of measurement, we often do not realize the important role being played by the science of metrology and metrological institutions in our day-to-day lives. Metrological systems play an important part in regulating our lives. For example, the regulatory aspects of weights and measures are part of a metrology system; the international metrology system makes it possible for one kilogram of a substance to be weighed identically in any part of the world; and the dissemination of time by metrology institutions helps us in planning and regulating schedules in our lives. International trade and commerce would not be possible in the absence of globally accepted measurement. Thus, measurements are not only important for metrologists working in industries and scientific laboratories but are also a matter of concern to all of us as consumers and citizens. There is a strong worldwide interest in the science of measurement and measurement results not only among metrologists but also among ordinary citizens. A general awareness is emerging of the need for reliable measurement results and associated issues.

METROLOGY AND TECHNOLOGICAL GROWTH

Socrates was an ancient Greek philosopher and thinker. His thoughts on philosophy, politics and the state, economics, and social systems have paved the way for subsequent thinkers to evolve new concepts, thoughts, and theories in these areas of learning. The quotation by Socrates at the beginning of this chapter states that measurements and numbering have come to the rescue of human understanding of nature. Before this, human beings were satisfied with comparisons, such as less or more, lighter or heavier, and longer or shorter. In historical perspective, this could be considered the beginning of the science of metrology in a systematic way, when human beings started measuring things and expressing the measurements in numbers. In the second quotation, David Layden defines measurement as a kind of language which ensures effective communication. He also defines the current concept of good measurement and the relationship between new measurement capabilities and the redefinition of the need for measurement. Between these two quotations lies the history of the evolution of metrology and metrological concepts.

The history of scientific advancement is closely linked with human capability to measure and measure accurately. The need for more and more accurate measurements increased with advances in various areas of science and technology. Scientists invented newer techniques of measurement to keep pace with these advances. This led to a better understanding of the nature and scientific principles associated with measurement. After the end of the Second World War, research in the basic sciences advanced at a rapid pace. This resulted in increased growth in high-technology industrial activity. The increasing complexity and speed of many modern processes and machines made automatic control essential, and such control was not possible without satisfactory means of taking measurements. The requirement for measurement capability increased in two ways. On the one hand, there has been an ever-increasing demand for measurement of new parameters, and on the other the need for more accurate measurements has increased, which was undreamed-of in the past. Spectacular advances in the technology of the measurement of physical quantities have been witnessed in the last few decades. The development of electrical sensors and the signal-processing capabilities of electronic circuits have extended the range of what can be measured. Thus, advances in metrology helped in the rapid growth of technology, and vice versa.

The harnessing of technology brought new types of devices and machines which affect human lives directly. It improved the quality of human lives but subjected them to certain types of technological risks. The discovery of electricity revolutionized the lifestyle of human beings but with an added risk of electric shock. Ecological imbalance and environmental pollution are direct results of the uncontrolled and unplanned use of technology. X-rays have useful applications in healthcare, but excessive exposure to them has a damaging effect on human beings. The microwave oven is a very useful domestic appliance, but overexposure to microwave radiation is a health hazard.

This harnessing of technology also brought another aspect of measurement into focus. The devices and machines developed as a result of advances in technology had to be safe for human use. Their harmful effects on human beings needed to be investigated and measured in quantitative terms. The safety criteria for these products had to be decided. The products were required to be evaluated so that the risk in using them was within acceptable limits of safety. This necessitated the development of reliable measurement techniques to study the effect of new technologies on human beings and to evaluate the safety of the products.

THE INCREASING NEED FOR RELIABLE MEASUREMENTS

A measurement can be considered reliable if the characteristics of the measurement result meet the objective for which the measurement is being made. Advances in technology necessitated matching advances in metrology. In the engineering industry, the tolerance for variation in manufactured parts had been decreasing by a factor of three each decade during the last few decades. As a result of these close tolerances, a marked improvement was observed in the efficiency and reliability of internal combustion engines for motor vehicles and gas turbines for aircraft. Such improvements depended on corresponding improvements in industrial metrology. In high-technology semiconductor production, the manufacturing tolerances are such that the required metrology is at the frontiers of physics. On the other hand, technological advances in metrology resulted in the development of sophisticated devices. With the invention of the atomic clock, the accuracy and resolution of time and frequency measurements were enhanced to an unbelievable degree. This helped in the development of very accurate navigational systems and long-distance measurements in radio astronomy.

Another area in which the need for accurate measurement is being felt is environmental monitoring. The adverse effect of industrial activity on the environment is becoming a matter of concern to humankind. Dangers are posed to human health and safety by environmental degradation and by the pesticide and heavy metal residues in the food we consume. Every new technology is resulting in new types of hazards. The qualitative and quantitative measurement of the hazards is a big challenge for analytical chemists.

Besides advances in technology and environmental monitoring, there are a number of other areas in which there is a need for reliable measurement. These include various facets of human endeavor in trade, industry, biological testing, forensic analysis, analysis of food and agricultural products, and technological testing of other products.

In the area of international trade, lack of acceptance of laboratory test data across national borders has been identified as a significant technical barrier to trade. Technical requirements for products are becoming more stringent. In order to have their products accepted internationally, manufacturers must be able to prove that their goods and services meet specified requirements. They must be able to show that laboratory test reports, inspection reports, product certificates, and system certificates conform to international standards. To achieve global acceptability of products and services, the metrological institutions in various countries have to ensure that the measurements they take are consistent and accurate to the desired level, as part of a global as well as national conformity assessment system.

METROLOGY CLASSIFICATIONS

An infinite number of measurements are made every day in trade, industry, and scientific laboratories throughout the world. However, the field of measurement can be classified in two ways. The first type of classification is based on the nature of the parameter being measured, and the second type is based on the application of the measurement result. A parameter under measurement can be either a physical parameter or an analytical or chemical parameter. This classification leads to two branches of metrology: physical metrology and chemical metrology. The application-based classification leads to three branches of metrology: legal metrology, industrial metrology, and scientific metrology. The classifications are shown in Figure 1.1.

Physical and Chemical Metrology

The science of physical measurement is called physical metrology, whereas the science of analytical and chemical measurement is called chemical metrology. Physical metrology deals with the measurement of physical parameters such as mass, length, electric current, velocity, and viscosity, while chemical metrology deals with the qualitative and quantitative analysis of substances used in the chemical, biological, medical, and environmental fields. The measurements made in metallurgical testing laboratories, pharmaceutical laboratories, and pathology laboratories fall under the purview of chemical metrology. Three broad areas—forensic analysis, environmental analysis, and analysis of food and agricultural products—are also part of chemical metrology.

Figure 1.1 Metrology classification.

Legal, Industrial, and Scientific Metrology

Irrespective of the physical or chemical nature of the measurements taken, the discipline of metrology may be classified into legal, industrial, and scientific branches depending upon the application of the measurement result.

Legal metrology

A credible measurement system is vital for trade in any society. All measurements related to trade and consumer protection come under the purview of legal metrology. Trade is a fundamental human activity, and it operates on the principle of fair exchange of products between two parties, which may be persons or organizations. Legal metrology ensures that all measurements made for the purpose of exchanging products as part of trade are fair and credible. This type of metrology is concerned with legal issues and activities such as the evaluation and certification of measuring instruments to ensure the accuracy, consistency, and credibility of measurements. Enforcing these issues is a function of the state. The instruments that are covered under legal metrology are regularly used to satisfy the multiple needs of society. Some representative examples of these instruments and related areas concerning human lives are given in Table 1.1.

Industrial metrology

Measurements are made in industries and service organizations to check the conformance of products and services to their specifications at various stages of product development, manufacturing, and servicing. These measurements come under the purview of industrial metrology.

Today’s mass production of quality goods and provision of quality services could not exist without an adequate measurement system. The requirements of the system for product and process monitoring have been stipulated in international standards such as the ISO 9000 series of quality management standards. Thus, an appropriate measurement system is an important resource for any manufacturing or service organization. In an industrial organization, measurements are required to be made at various stages in the life cycles of products, which are as follows:

Product development: This stage includes the design and development of the product as well as pilot manufacturing. The prototypes of products are required to be evaluated and tested to ensure conformance to their specifications. The specifications are drawn up based on the requirements of customers.

Inspection of raw material and components: Measurements are made with a view to checking the conformance of the raw material and components to their specifications by testing and evaluation. This is called incoming inspection, and it ensures that the end product meets the requirements of its specifications.

Product manufacturing: At various stages of the manufacturing process, the product is inspected before it goes to the next stage. This is called in-process inspection. Again, the end product is tested to check conformance to its specifications. This is also called final inspection. At various inspection stages, measurements are made to determine the product’s characteristics.

Process monitoring: Statistical process control (SPC) techniques are extensively used by the industry to control the manufacturing process and to ensure that product quality is within the stipulated specification limits. Measurements at specified regular intervals are made of certain characteristics of the product drawn from the process, and key parameters derived based on the measurement results are plotted on a control chart. Average and range charts are very popular control charts used by the industry. After the results are plotted on the control chart, a decision is made as to whether the manufacturing process is under statistical control or not. If the process is found to be out of control, corrective measures are taken to bring the process under statistical control to ensure that the process is capable of manufacturing products of the desired quality.

Product maintenance: During the process of corrective maintenance, the measurement results serve as diagnostic tools for finding the causes of product unserviceability. Based on the diagnosis, corrective maintenance action is completed. Measurement results are also used in preventive maintenance.

Scientific metrology

Measurements made in scientific laboratories are categorized as scientific metrology. Following are some examples of such measurements.

Measurements made in testing laboratories: A product is tested in a testing laboratory with a view to determining its characteristics and deciding whether the product complies with its specifications. The product may be an industrial, agricultural, or natural product. The pathological and analytical laboratories used for diagnosis of human ailments are one type of testing laboratory. Forensic science laboratories for the solving of crimes are another type. Today, with advancements in technology, new types of products and materials are being designed and manufactured. The need to test these products is resulting in a need for newer types of test methods and testing techniques. A testing laboratory can be a captive laboratory set up by an organization to meet its testing needs. It can also be a third-party independent test laboratory that provides testing services to its clients. Testing activities are always associated with physical or chemical measurements.

Measurements made in research and development (R&D) laboratories: Research activity is linked with the innovative capability of the human mind. Certain theories and hypotheses are proposed in the areas of the basic sciences and technology. Measurements are made in R&D laboratories to verify these theories. The correctness or otherwise of a proposed theory is decided based upon analysis of the measurement results obtained. For R&D laboratories involved in product research and development, measurement results help the scientists conclude whether the desired characteristics of a product have been achieved.

There is an interesting anecdote about the experimental verification of Einstein’s general theory of relativity. Einstein predicted that light should be bent by gravitational fields. Thus, when the light from a distant star passes close to the sun, it should be deflected due to the gravitational field of the sun, causing the star to appear in a different position to an observer on earth. It is difficult to see this effect, as the light from the sun makes it impossible to observe stars that appear near the sun in the sky. However, it is possible to do so during a period of total solar eclipse. To verify the postulates of Einstein’s theory, experiments were conducted by British scientists in 1919 in West Africa during a period of total solar eclipse. The experiments showed that light was deflected by the sun as predicted by Einstein’s theory. This proved the correctness of the theory. However, later examination of the photographs taken during the experiment showed that the errors were as great as the effect the scientists were trying to measure. The success of the experiment was thus sheer luck. The light deflection was in fact accurately measured in a number of experiments carried out later.

Measurements made in calibration laboratories: All measurements are made using some type of testing and measuring instruments. Analytical measurements also involve measuring instruments, standard chemicals, and certified reference materials. A measurement result is valid only if it has been made with a measuring instrument which has been calibrated. Calibration is the process of ensuring the integrity of measuring instruments. The behavior of a measuring instrument is compared against a duly calibrated laboratory reference standard. Based on a comparison of the measurement results obtained by the instrument being calibrated with those obtained by a duly calibrated reference standard, the operational integrity of the measuring instrument is ascertained and certified. This service is provided by calibration laboratories. The process of calibration also ensures the traceability of measurement results to national standards. The concepts pertaining to calibration and traceability are explained in detail in subsequent chapters.

THE PROBABILISTIC NATURE OF MEASUREMENT RESULTS

A measurement result is a quantitative statement. It is always expressed as a number followed by the unit of measurement. The unit should be the commonly accepted unit of the parameter being measured. Though the measurement result is expressed as a number, its nature is very different from that of an arithmetic number. An arithmetic number is an exact number. If we refer to the arithmetic number 3.85, there is no ambiguity in anybody’s mind about this number’s being 3.85. On the other hand, if we express a measurement result as 3.85, many questions may come into the minds of the person making the measurement and the persons using the measurement result. For example, if a measurement result indicates that a particular device consumes an electric current of 3.85 amperes, the following questions may come to the mind of the person making the measurement:

If the measuring device had lower resolution, the measurement result could have been 3.8 amperes or 3.9 amperes

If the measuring device had better resolution, the result could have been 3.854, 3.853, or another similar number

If another person had made the measurement, a slightly different result could have been obtained

The consequence of all these ifs is that the measurement result might not have been the same. At the same time, they would also not have been very different from 3.85 amperes. It is thus an accepted situation that repeat measurement results will not be the same but will be similar, that is, very close to 3.85 amperes. These repeat measurements have a common origin, and hence there is a commonality among them. As repeat measurements yield slightly different results, there is always a shadow of a doubt concerning the credibility of a measurement result. The greater the spread of repeat measurement results, the greater the shadow of a doubt. It is therefore an accepted fact that repeat measurements of the same parameter obtained from any measurement system will have a certain amount of variability. The level of variability will be different depending upon the technological sophistication of the measurement system. This variability among different values of repeat measurement results leads to confusion as to what the actual value of the parameter being measured is. We therefore conclude that there is always error associated with a measurement result. As regards the reason for this variability, it is considered to be the cumulative effect of the variations in a number of factors that are influencing the measurements. Some of the factors are the inherent limitation of the measurement method, variation in the repeat observations of measuring instruments in the measurement setup, and the technical competence of the person making the measurements.

Thus, repeat measurements are spread around a specific value. This spread represents an interval. The measurement result is expected to lie within this interval most of the time. This spread characterizes the uncertainty of measurement results, the so-called shadow of a doubt. The less the measurement uncertainty, the less the shadow of a doubt, and the more the uncertainty, the greater the doubt. It should therefore be clear that a measurement result of a specific parameter obtained from a measurement system is one of the numbers contained in the interval. It is therefore not an exact number, or the number, as an arithmetic number is. In pure mathematics, the value of the square root of 2 or of n can be given accurately. The value can be exactly predicted to any number of decimal points. Occurrences which can be predicted accurately in this way are called deterministic occurrences. The concept of determinism implies that once an occurrence is known, successive occurrences can be predicted. Measurement systems are governed by scientific laws which are not deterministic in nature. It is never possible to predict accurately in advance successive measurement results from previous experience, but one can roughly predict a range of values between which measurement results can be expected to lie. These concepts have been clearly defined by the Nobel laureate Max Planck in the following words:

Simple as we make the conditions and precise as our measuring instruments may be, we shall never succeed in calculating in advance the results of actual measurements with an absolute accuracy, in contrast to calculations in pure mathematics as in the case of the square root of 2 or of n, which can be given accurately to any number of decimal places. (Scientific Autobiography, 1947)

Thus measurement results are indeterministic rather than deterministic. This implies that measurement results are not exact values. They follow certain statistical laws and are probabilistic in nature. There are a large number of values within an interval that can claim to represent the value of the parameter. These values are not the same but are similar. There is a certain probability that the repeat measurement results will lie in this interval most of the time. This spread of measurement results is a measure of the indecisiveness of the measurement system. This indecisiveness is called measurement uncertainty and is the subject matter of this book.

WHY TAKE MEASUREMENTS?

There is a purpose for every measurement. Measurement results are used as tools for decision making. Irrespective of the nature of a measurement, it is always followed by a decision. The decision can be a conscious or a subconscious one. With measurements that are involved in trade, the trading partners are ensuring that they are both treated fairly. In an industrial scenario, measurement results are used to make decisions on product acceptance. If a product conforms to its specifications, it is accepted; otherwise, it is rejected. Measurement results are also used to improve the quality of products or processes. SPC techniques are used in industry with a view to controlling manufacturing processes. Control charts are drawn based on the measurement results of sample items picked from the production line at specified intervals. The study of these charts tells us whether a process is under statistical control or not. The determination to be made here is whether the manufacturing process is under statistical control and should be left undisturbed or whether it is out of control and needs corrective action.

Similarly, in a testing laboratory, measurement results help us conclude whether the product characteristics are as required. In a calibration laboratory, measurement results help us decide whether measuring equipment is capable of measuring a particular parameter to the desired accuracy. In an R&D laboratory, measurement results help us verify the correctness of a proposed scientific concept. The analytical results obtained in a forensic science laboratory help in deciding cases in a court of law. Similarly, the analytical results obtained in a pathology laboratory helps doctors decide whether an individual is suffering from a suspected disease.

Thus, the purpose of measurement is to obtain information on entities and make decisions based on the information contained in the measurement result. Decisions are made based on comparison of the numerical value of the measurement result with specified criteria. The criteria are also called specification limits. Specification limits are determined based on the intended end use of an item. For example, the specification limits for the dimensions of a manufactured part and for the lower limit of voltage that an electric cable should withstand fall into this category. In certain situations, specification limits are also decided based on experience or on the characteristics of similar high-quality products. The decision criteria for body fluid composition are an example of this category.

There are two ways of stipulating specification limits, depending upon the requirements and the decisions to be made:

One-way specification limits: The decision criterion is a single value of the parameter. A decision is made based on whether the measurement result is smaller or larger than the specified value. One example is the decision on the quality of a microwave oven based on measurement of the leakage of microwave radiation from the oven. If the measured value is less than a specified value, it is decided that the design of the microwave oven is acceptable and that it is safe to use the oven. The specification limit is derived based on expert studies of the effects of microwave radiation on the human body. Similarly, if the tensile strength of an iron wire is greater than a specified value, a decision is made that it can be used in armoring electric cables for underground installation. A person can be charged for a traffic offense if the speed of the vehicle driven by the person as measured by police radar is greater than a specified value. The specification limit in this case depends upon factors such as traffic density and the type of road.

Two-way specification limits: In this case, a decision is made based on two specified values. They are called the upper specification limit (USL) and the lower specification limit (LSL). The measurement result is compared against these two limits, and a decision is made accordingly. For example, if the measured dimensions of a manufactured automobile part are within the two specification limits, a decision to accept the part is made; otherwise, a decision to reject it is made. In calibration laboratories, measuring equipment is categorized as acceptable if measurement results show that the equipment reads the value of the reference standard within acceptable limits of accuracy. The decision criteria are shown in Figure 1.2.

Figure 1.2 Decision criteria for two-way specification limits.

RISKS ASSOCIATED WITH STATISTICAL DECISION MAKING

Due to the probabilistic nature of measurement results, there is always uncertainty regarding the measured value of a parameter. This uncertainty is reflected in the correctness of the decision made based on a measurement result. This is especially true when the measurement result is close to the specification limits, where there is more likelihood of the decision’s being incorrect. This is illustrated by the following example.

Product: Electric cable

Parameter being measured: Insulation resistance

Specification limits: Greater than 10 megaohms at 500 volts DC

Decision criterion: Measure the insulation resistance of the electric cable at 500 volts DC; cable to be accepted if insulation resistance is greater than 10 megaohms, otherwise rejected

Suppose the value of the parameter—the insulation resistance of the cable—is 10.6 megaohms: the cable should be accepted, since the value of the parameter is greater than the specification limit of 10 megaohms. This is an ideal situation. However, the actual measurement result will not always be 10.6 megaohms. Because of the probabilistic nature of repeat measurements, the measured values will lie within an interval around 10.6 megaohms most of the time. If the interval is indicated by 10.6 ±1.0 megaohms, the measurement results will lie between 9.6 megaohms and 11.6 megaohms. If repeat measurements are made on the same cable, the cable will be rejected when the measured value lies between 9.6 and 10.0 megaohms. It will be accepted when the measured value lies between 10.0 and 11.6 megaohms. If the measurements are spread uniformly within the above interval, 20% of the time the measured value will be between 9.6 and 10.0 megaohms, and 80% of the time between 10.0 and 11.6 megaohms. Thus, 20% of the time the cable will be wrongly rejected because of the uncertainty of the measurement result. This situation is illustrated in Figure 1.3.

Thus, because of measurement uncertainty, it may happen that a measured value is observed to be outside the specification limits whereas the value of the parameter is actually within the specification limits. It could also be that the actual value of the parameter is outside the specification limits but the measured value is within the specification limits. The rejection of a conforming item or